Env setupu¶

In [1]:
import scanpy as sc, anndata as ad, numpy as np, pandas as pd
from scipy import sparse
import matplotlib.pyplot as plt
import plotly.express as px
from anndata import AnnData
from sklearn.neighbors import KernelDensity
import random
from matplotlib import pylab
from matplotlib.colors import TwoSlopeNorm

import warnings
import scanpy.external as sce 
import yaml

import seaborn as sns
import os
import sys

warnings.filterwarnings('ignore')
In [2]:
import anndata2ri
import rpy2.rinterface_lib.callbacks
import logging
In [3]:
rpy2.rinterface_lib.callbacks.logger.setLevel(logging.ERROR)
In [4]:
anndata2ri.activate()
In [5]:
%load_ext rpy2.ipython
In [6]:
sc.settings.verbosity = 3  
sc.logging.print_header()
nprocs=4
np.random.seed(0)
with open("../data/resources/rcParams.yaml") as f:
    rcParamsDict = yaml.full_load(f)
    for k in rcParamsDict["rcParams"]:
        print("{} {}".format(k,rcParamsDict["rcParams"][k]))
        plt.rcParams[k] = rcParamsDict["rcParams"][k]
    for k1 in set(list(rcParamsDict)).difference(set(["rcParams"])):
        print("{} {}".format(k1,rcParamsDict[k1]))

scanpy==1.9.3 anndata==0.8.0 umap==0.5.1 numpy==1.20.3 scipy==1.10.1 pandas==2.0.3 scikit-learn==1.3.2 statsmodels==0.13.5 python-igraph==0.9.1 louvain==0.7.0 leidenalg==0.8.3 pynndescent==0.5.12
figure.dpi 80
savefig.dpi 500
figure.figsize [10, 10]
axes.facecolor None
figure.facecolor None
dotSize 20

Configure paths¶

In [7]:
outdir = "../data/output"

with open("../data/resources/iPSC_lines_map.yaml", 'r') as f:
    iPSC_lines_map = yaml.load(f, Loader=yaml.FullLoader)["lines"]
colorsmap = dict(zip([i["newName"] for i in iPSC_lines_map.values()],[i["color"] for i in iPSC_lines_map.values()]))


figDir = "./figures"
if not os.path.exists(figDir):
   # Create a new directory because it does not exist
   os.makedirs(figDir)


figDir = "./figures"
if not os.path.exists(figDir):
   # Create a new directory because it does not exist
   os.makedirs(figDir)
    
#tradeSeqUtilsPath = "/opt/tradeSeq_utils.R"
tradeSeqUtilsPath = "/opt/tradeSeq_utils.R"

var setting¶

In [8]:
branchTag = "Astro"
endpoint = "endpoint_OuterRadialGliaAstrocytes"
markers = ["RELN","PAX6","CBLN1","MKI67"]
min_cells = 50
MultiTPTstages = ["early","mid","late"]
binWidthRate = .05
ResamplingIterations = 50
leidenOrder = ["ProliferatingProgenitors", 'RadialGliaProgenitors', 'OuterRadialGliaAstrocytes']
In [9]:
BranchBCs = sc.read_h5ad(outdir+"/adatas/adataPaga.h5ad")
sc.pl.draw_graph(BranchBCs[BranchBCs.obs[endpoint] == 1], color = ["cellID_newName","leidenAnnotated","dataset"], size = rcParamsDict["dotSize"],add_outline=True, outline_width=(0.2, 0.05), vmin='p2',vmax='p98', wspace=.4, ncols=2,legend_fontsize='xx-large')
BranchBCs = BranchBCs.obs_names[BranchBCs.obs[endpoint] == 1].tolist()

harmonyTheta = 1
HVGs=pd.read_csv(outdir+"/HVG_list_intersection_Curated_"+branchTag+"_PagaTest.txt", sep = "\t")["HVG"]

Branch specifi pre-processing¶

In [10]:
adataBranch = sc.read_h5ad(outdir+'/adatas/MultiplexingPreprocessing_unscaled.h5ad')[BranchBCs]
adataBranch.obs["leidenAnnotated"] =  sc.read_h5ad(outdir+"/adatas/ClusterAnnotated_Base_filt.h5ad")[adataBranch.obs_names].obs["leidenAnnotated"]
adataBranch.uns["leidenAnnotated_colors"] =  sc.read_h5ad(outdir+"/adatas/adataPaga.h5ad")[adataBranch.obs_names].uns["leidenAnnotated_colors"]
goodcls = adataBranch.obs.groupby(["leidenAnnotated"]).size()[adataBranch.obs.groupby(["leidenAnnotated"]).size() >= 50].index.tolist()
adataBranch = adataBranch[adataBranch.obs["leidenAnnotated"].isin(goodcls)]


adataBranch.layers["unscaled"] = adataBranch.X.copy()


sc.pp.normalize_total(adataBranch, target_sum=1e4)

sc.pp.log1p(adataBranch)
adataBranch.raw = adataBranch



adataBranch = adataBranch[:,HVGs]
adataBranch.var["highly_variable"] = True
#sc.pp.highly_variable_genes(adataBranch, min_mean=0.0125, max_mean=5, min_disp=0.5, batch_key = "dataset")

sc.pp.regress_out(adataBranch, ['total_counts','pct_counts_mt'], n_jobs=4)
sc.pp.scale(adataBranch, zero_center=False)
#adataBranch.X = adataBranch.X + np.abs(adataBranch.X.min())

normalizing counts per cell
    finished (0:00:00)
regressing out ['total_counts', 'pct_counts_mt']
    sparse input is densified and may lead to high memory use
    finished (0:02:21)
In [11]:
sc.tl.pca(adataBranch, svd_solver='arpack', use_highly_variable=True)
sce.pp.harmony_integrate(adataBranch, 'dataset', max_iter_harmony=20,adjusted_basis="X_pca",random_state = 0, theta = harmonyTheta, lamb = 2)
sc.pp.neighbors(adataBranch,use_rep="X_pca",n_neighbors=50, n_pcs=10, random_state=0)

sc.tl.umap(adataBranch, random_state=0)

computing PCA
    on highly variable genes
    with n_comps=50
    finished (0:00:04)

2024-06-04 10:21:10,230 - harmonypy - INFO - Iteration 1 of 20
2024-06-04 10:21:12,939 - harmonypy - INFO - Iteration 2 of 20
2024-06-04 10:21:15,662 - harmonypy - INFO - Converged after 2 iterations

computing neighbors
    finished: added to `.uns['neighbors']`
    `.obsp['distances']`, distances for each pair of neighbors
    `.obsp['connectivities']`, weighted adjacency matrix (0:00:11)
computing UMAP
    finished: added
    'X_umap', UMAP coordinates (adata.obsm) (0:00:28)
In [12]:
sc.pl.pca(adataBranch, color = ["dataset","stage",'leidenAnnotated'], size = 20,add_outline=True, outline_width=(0.2, 0.05), vmin='p2',vmax='p98',wspace=.4)

DPT¶

In [13]:
sc.tl.diffmap(adataBranch)


RootName = adataBranch.obs_names[adataBranch[:, ["TOP2A"]].X.argmax()]
root_idx = np.where(adataBranch.obs_names == RootName)[0][0]
adataBranch.uns['iroot'] = root_idx


sc.tl.dpt(adataBranch)
sc.pl.diffmap(adataBranch, color = ["dpt_pseudotime","TOP2A"],size=rcParamsDict["dotSize"],add_outline=True, outline_width=(0.2, 0.05), vmin='p2',vmax='p98', legend_fontsize="xx-large", frameon=False, wspace=.3, legend_loc="on data")

computing Diffusion Maps using n_comps=15(=n_dcs)
computing transitions
    finished (0:00:00)
    eigenvalues of transition matrix
    [1.         0.9874326  0.97819453 0.9609501  0.955528   0.94372886
     0.94041723 0.93345386 0.9252512  0.9154983  0.90268844 0.89372355
     0.88754183 0.88612014 0.8749523 ]
    finished: added
    'X_diffmap', diffmap coordinates (adata.obsm)
    'diffmap_evals', eigenvalues of transition matrix (adata.uns) (0:00:00)
computing Diffusion Pseudotime using n_dcs=10
    finished: added
    'dpt_pseudotime', the pseudotime (adata.obs) (0:00:00)
In [14]:
adataBranch.write_h5ad(outdir+"/adatas/AdataLineage."+branchTag+".h5ad")
In [15]:
timeCov="dpt_pseudotime"
from matplotlib.pyplot import figure
sc.settings.set_figure_params(dpi=100)
pylab.rcParams['figure.figsize'] = (10, 8)

By timepoint¶

In [16]:
sns.set_style("white")
In [17]:
cov = "stage"
In [18]:
#errors by dataset
for Stage in adataBranch.obs["stage"].unique(): 
    StageLT = adataBranch.obs[adataBranch.obs["stage"] == Stage]
    
    StageLT_smoothed = pd.DataFrame(index=np.linspace(0, 1, 10000))
    for dataset in StageLT["dataset"].unique():
        datasetLT = StageLT.loc[StageLT["dataset"] == dataset,timeCov]
        _bandwidth = round((datasetLT.max() - datasetLT.min()) * binWidthRate, 2)
    
        KernelData = datasetLT

        
        X = KernelData.values[:, np.newaxis]
        X_plot = np.linspace(0, 1, 10000)[:, np.newaxis]

        # Pick bandwidht
        kde_Calc = KernelDensity(kernel="gaussian", bandwidth=_bandwidth).fit(X)
        Y=np.exp(kde_Calc.score_samples(X_plot))
        
        StageLT_smoothed[dataset] = Y
        
    StageLT_smoothed["dpt_cut"] = StageLT_smoothed.index.tolist()
        
    # Calculate error and mean on smoothe values ()

    StageLT_smoothed["mean"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].mean(axis =1)
    StageLT_smoothed["stdev"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].std(axis =1)
    StageLT_smoothed["errMin"] = StageLT_smoothed["mean"] - StageLT_smoothed["stdev"]
    StageLT_smoothed["errMax"] = StageLT_smoothed["mean"] + StageLT_smoothed["stdev"]
    StageLT_smoothed["errMin"] = np.where(StageLT_smoothed["errMin"] < 0, 0, StageLT_smoothed["errMin"])
    
    plt.plot(StageLT_smoothed["dpt_cut"], StageLT_smoothed["mean"], '-', label =Stage,linewidth = 3,color = dict(zip(adataBranch.obs[cov].cat.categories.tolist(),adataBranch.uns["stage_colors"]))[Stage])
    plt.fill_between(StageLT_smoothed["dpt_cut"], StageLT_smoothed["errMin"], StageLT_smoothed["errMax"] , alpha=0.2,color = dict(zip(adataBranch.obs[cov].cat.categories.tolist(),adataBranch.uns["stage_colors"]))[Stage])
    plt.xlabel('pseudotime', size=30)
    plt.xticks(size=30)
    plt.yticks(size=30)

    plt.ylabel('Cells density', size=30)
    plt.legend(prop={'size': 30})
    plt.savefig(figDir+"/8_LineageFigures.dpt."+branchTag+".svg")
In [19]:
fig, ax  = plt.subplots(1,1, figsize=(20,10))

adataBranchCTL02 = adataBranch[adataBranch.obs["cellID_newName"] == "CTL02A"]
for Stage in adataBranchCTL02.obs["stage"].unique(): 
    StageLT = adataBranchCTL02.obs[adataBranchCTL02.obs["stage"] == Stage]
    
    StageLT_smoothed = pd.DataFrame(index=np.linspace(0, 1, 10000))
    for dataset in StageLT["dataset"].unique():
        datasetLT = StageLT.loc[StageLT["dataset"] == dataset,timeCov]
        _bandwidth = round((datasetLT.max() - datasetLT.min()) * binWidthRate, 2)
        KernelData = datasetLT
        
        
        X = KernelData.values[:, np.newaxis]
        X_plot = np.linspace(0, 1, 10000)[:, np.newaxis]

        # Pick bandwidht
        kde_Calc = KernelDensity(kernel="gaussian", bandwidth=_bandwidth).fit(X)
        Y=np.exp(kde_Calc.score_samples(X_plot))
        
        StageLT_smoothed[dataset] = Y
        
    StageLT_smoothed["dpt_cut"] = StageLT_smoothed.index.tolist()
        
    # Calculate error and mean on smoothe values ()

    StageLT_smoothed["mean"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].mean(axis =1)
    StageLT_smoothed["stdev"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].std(axis =1)
    StageLT_smoothed["errMin"] = StageLT_smoothed["mean"] - StageLT_smoothed["stdev"]
    StageLT_smoothed["errMax"] = StageLT_smoothed["mean"] + StageLT_smoothed["stdev"]
    StageLT_smoothed["errMin"] = np.where(StageLT_smoothed["errMin"] < 0, 0, StageLT_smoothed["errMin"])
    
    plt.plot(StageLT_smoothed["dpt_cut"], StageLT_smoothed["mean"], '-', label =Stage,linewidth = 3,color = dict(zip(adataBranchCTL02.obs[cov].cat.categories.tolist(),adataBranchCTL02.uns["stage_colors"]))[Stage])
    plt.fill_between(StageLT_smoothed["dpt_cut"], StageLT_smoothed["errMin"], StageLT_smoothed["errMax"] , alpha=0.2,color = dict(zip(adataBranchCTL02.obs[cov].cat.categories.tolist(),adataBranchCTL02.uns["stage_colors"]))[Stage])
    plt.xlabel('pseudotime', size=30)
    plt.xticks(size=30)
    plt.yticks(size=30)

    plt.ylabel('Cells density', size=30)
    plt.legend(prop={'size': 30})
    plt.savefig(figDir+"/8_LineageFigures.dpt.CTL02A."+branchTag+".svg")

By paradigm¶

In [19]:
cov = "type"
In [20]:
#with errors by dataset
for Stage in adataBranch.obs["type"].unique(): 
    StageLT = adataBranch.obs[adataBranch.obs["type"] == Stage]
    
    StageLT_smoothed = pd.DataFrame(index=np.linspace(0, 1, 10000))
    for dataset in StageLT["dataset"].unique():
        datasetLT = StageLT.loc[StageLT["dataset"] == dataset,timeCov]
        _bandwidth = round((datasetLT.max() - datasetLT.min()) * binWidthRate, 2)
    
        KernelData = datasetLT

        
        X = KernelData.values[:, np.newaxis]
        X_plot = np.linspace(0, 1, 10000)[:, np.newaxis]

        # Pick bandwidht
        kde_Calc = KernelDensity(kernel="gaussian", bandwidth=_bandwidth).fit(X)
        Y=np.exp(kde_Calc.score_samples(X_plot))
        
        StageLT_smoothed[dataset] = Y
        
    StageLT_smoothed["dpt_cut"] = StageLT_smoothed.index.tolist()
        
    # Calculate error and mean on smoothe values ()

    StageLT_smoothed["mean"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].mean(axis =1)
    StageLT_smoothed["stdev"] = StageLT_smoothed[[i for i in StageLT_smoothed.columns if i not in ['dpt_cut','stdev']]].std(axis =1)
    StageLT_smoothed["errMin"] = StageLT_smoothed["mean"] - StageLT_smoothed["stdev"]
    StageLT_smoothed["errMax"] = StageLT_smoothed["mean"] + StageLT_smoothed["stdev"]
    StageLT_smoothed["errMin"] = np.where(StageLT_smoothed["errMin"] < 0, 0, StageLT_smoothed["errMin"])
    
    plt.plot(StageLT_smoothed["dpt_cut"], StageLT_smoothed["mean"], '-', label =Stage,linewidth = 3,color = dict(zip(adataBranch.obs[cov].cat.categories.tolist(),adataBranch.uns["type_colors"]))[Stage])
    plt.fill_between(StageLT_smoothed["dpt_cut"], StageLT_smoothed["errMin"], StageLT_smoothed["errMax"] , alpha=0.2,color = dict(zip(adataBranch.obs[cov].cat.categories.tolist(),adataBranch.uns["type_colors"]))[Stage])

    plt.xlabel('pseudotime', size=30)
    plt.xticks(size=30)
    plt.yticks(size=30)

    plt.ylabel('Cells density', size=30)
    plt.legend(prop={'size': 30})
    plt.savefig(figDir+"/8_LineageFigures.Up_vs_Down.dpt."+branchTag+".svg")

Explore Genotypes¶

In [21]:
#######---original adata ------- Load Original adata and plot uncorrected PCA and original compositions

sc.settings.verbosity = 0 


adataInitPCA = adataBranch.copy()


compositions = pd.DataFrame(adataInitPCA.obs.groupby(["stage","cellID_newName"]).size())
compositions = compositions.reset_index().rename(columns={0:"number_of_cells"})
fig = px.bar(compositions, x="stage", y="number_of_cells", color="cellID_newName", 
             category_orders={"stage":["early","mid","late"]}, height=800,width=1000, template="plotly_white",
             color_discrete_map = dict(zip(adataInitPCA.obs["cellID_newName"].cat.categories, adataInitPCA.uns["cellID_newName_colors"])))

fig.update_traces(marker_line_color='rgb(8,48,107)',
                  marker_line_width=1.5, opacity=1)

fig.show()

Downsampling loops¶

In [22]:
ReproducibilityDF = pd.DataFrame() 
ReproducibilityRaw = pd.DataFrame() 


adataInitPCA = adataInitPCA[adataInitPCA.obs["stage"].isin(MultiTPTstages)]
#Remove genotypes not in all timepoints
GoodIDs = pd.crosstab(adataInitPCA.obs.cellID_newName,adataInitPCA.obs.stage ) > min_cells
GoodIDs = GoodIDs[GoodIDs.sum(axis =1) == len(adataInitPCA.obs["stage"].unique())].index.tolist()
adataInitPCA = adataInitPCA[adataInitPCA.obs.cellID_newName.isin(GoodIDs)]
    
compositions = pd.DataFrame(adataInitPCA.obs.groupby(["stage","cellID_newName"]).size())
compositions = compositions.reset_index().rename(columns={0:"number_of_cells"})
fig = px.bar(compositions, x="stage", y="number_of_cells", color="cellID_newName", title="Balance Before each loop",
             category_orders={"stage":["early","mid","late"]}, height=800,width=1000, template="plotly_white",
             color_discrete_map = dict(zip(adataInitPCA.obs["cellID_newName"].cat.categories, adataInitPCA.uns["cellID_newName_colors"])))

fig.update_traces(marker_line_color='rgb(8,48,107)',
                  marker_line_width=1.5, opacity=1)

fig.show()

    
for seed in range(ResamplingIterations):
    random.seed(seed)
    
    adataGenotypes = adataInitPCA.copy()


    #Subset timepoint if needed
    MinCells = pd.crosstab(adataGenotypes.obs.cellID_newName,adataGenotypes.obs.stage ) 
    MinCells = MinCells.min().to_dict()

    sampledBCs = []
    for stage in adataGenotypes.obs.stage.unique():
        samplingSize = MinCells[stage]
        sampledBCsStage = [ random.sample(adataGenotypes[(adataGenotypes.obs["stage"] == stage) & (adataGenotypes.obs["cellID_newName"] == ID)].obs_names.tolist() ,samplingSize) for ID in adataGenotypes.obs["cellID_newName"].unique().tolist()  ]
        sampledBCsStage = [item for sublist in sampledBCsStage for item in sublist]
        sampledBCs = sampledBCs + sampledBCsStage

    adataGenotypes = adataGenotypes[sampledBCs]
    
    #Trim extreme values
#     trimExtremes = [0.01,.99]
#     minDPT, maxDPT = np.quantile(adataGenotypes.obs["pc1"], trimExtremes)
#     adataGenotypes = adataGenotypes[(adataGenotypes.obs["pc1"] > minDPT) & (adataGenotypes.obs["pc1"] < maxDPT)]
    
    _bandwidth = round((adataGenotypes.obs["dpt_pseudotime"].max() - adataGenotypes.obs["dpt_pseudotime"].min()) * binWidthRate, 2)
    seedPseudotime = pd.DataFrame()
    seedPseudotimeRaw = pd.DataFrame()

    for cellID in adataGenotypes.obs["cellID_newName"].unique(): 
        LocalPseudotime = pd.DataFrame()
        LocalPseudotimeRaw = pd.DataFrame()
        KernelData = adataGenotypes.obs.loc[adataBranch.obs["cellID_newName"] == cellID, "dpt_pseudotime"]
        X = KernelData.values[:, np.newaxis]
        X_plot = np.linspace(adataGenotypes.obs["dpt_pseudotime"].min(), adataGenotypes.obs["dpt_pseudotime"].max(), 10000)[:, np.newaxis]
        # Pick bandwidht
        kde_Calc = KernelDensity(kernel="gaussian", bandwidth=_bandwidth).fit(X)
        Y=np.exp(kde_Calc.score_samples(X_plot))
        LocalPseudotime[seed] = Y
        LocalPseudotime["cellID"] = cellID
        LocalPseudotime["x"] = X_plot.flatten()
        seedPseudotime = pd.concat([seedPseudotime,LocalPseudotime ], ignore_index=True, axis = 0)
        #Raw values non smoothed
        LocalPseudotimeRaw[seed] = KernelData.tolist()
        LocalPseudotimeRaw["cellID"] = cellID
        seedPseudotimeRaw = pd.concat([seedPseudotimeRaw,LocalPseudotimeRaw], ignore_index=True)
    if "cellID" in ReproducibilityDF.columns:
        ReproducibilityDF = pd.concat([seedPseudotime[seed], ReproducibilityDF], axis = 1)
        ReproducibilityRaw = pd.concat([seedPseudotimeRaw[seed], ReproducibilityRaw], axis = 1)
    else:
        ReproducibilityDF = pd.concat([ReproducibilityDF, seedPseudotime], axis =1 )
        ReproducibilityRaw = pd.concat([ReproducibilityRaw, seedPseudotimeRaw], axis =1 )
 
        
        
compositions = pd.DataFrame(adataGenotypes.obs.groupby(["stage","cellID_newName"]).size())
compositions = compositions.reset_index().rename(columns={0:"number_of_cells"})
fig = px.bar(compositions, x="stage", y="number_of_cells", color="cellID_newName", title="Balance Operated whithin each loop",
             category_orders={"stage":["early","mid","late"]}, height=800,width=1000, template="plotly_white",
             color_discrete_map = dict(zip(adataGenotypes.obs["cellID_newName"].cat.categories, adataGenotypes.uns["cellID_newName_colors"])))

fig.update_traces(marker_line_color='rgb(8,48,107)',
                  marker_line_width=1.5, opacity=1)

fig.show()
In [23]:
plotDF = ReproducibilityDF.copy()

plotDF["std"] = plotDF[[c for c in plotDF.columns if c not in ["x","cellID","std","mean",'max','min']]].std(axis =1)

plotDF["mean"] = plotDF[[c for c in plotDF.columns if c not in ["x","cellID","std","mean",'max','min']]].mean(axis =1)
plotDF["max"] = plotDF["mean"]+plotDF["std"]
plotDF["min"] = plotDF["mean"]-plotDF["std"]

for g in plotDF.cellID.unique():
    plotDFID = plotDF[plotDF.cellID == g]
    plt.plot(plotDFID["x"], plotDFID["mean"], '-',
             label =g,
             color = dict(zip(adataGenotypes.obs["cellID_newName"].cat.categories.tolist(),adataGenotypes.uns["cellID_newName_colors"]))[g],
             linewidth = 3)
    plt.fill_between(plotDFID["x"], plotDFID["min"], plotDFID["max"] , 
                     alpha=0.2,label=g,
                     color = dict(zip(adataGenotypes.obs["cellID_newName"].cat.categories.tolist(),adataGenotypes.uns["cellID_newName_colors"]))[g])

    plt.xlabel('pseudotime', size=30)
    plt.xticks(size=30)
    plt.yticks(size=30)
    plt.legend(prop={'size': 30})
    plt.savefig(figDir+"/8_LineageFigures.Genotypes."+branchTag+".svg")
In [43]:
DPTdf = plotDF.rename(columns=dict(zip([c for c in plotDF.columns if isinstance(c, int)], ["iteration."+str(c) for c in plotDF.columns if isinstance(c, int)]))).rename(columns={"x":"pseudotime_bin"})
DPTdf["Branch"] = branchTag
DPTdf.to_csv(outdir+"/PseudotimeGenotypeDensity."+branchTag+".tsv", index=False, sep="\t")

Tradeseq within lineage¶

In [24]:
plt.rcParams.update(plt.rcParamsDefault)
with open("../data/resources/rcParams.yaml") as f:
    rcParamsDict = yaml.full_load(f)
    for k in rcParamsDict["rcParams"]:
        print("{} {}".format(k,rcParamsDict["rcParams"][k]))
        plt.rcParams[k] = rcParamsDict["rcParams"][k]
    for k1 in set(list(rcParamsDict)).difference(set(["rcParams"])):
        print("{} {}".format(k1,rcParamsDict[k1]))

figure.dpi 80
savefig.dpi 500
figure.figsize [10, 10]
axes.facecolor None
figure.facecolor None
dotSize 20
In [25]:
obs = adataBranch.obs
obsm = adataBranch.obsm


adataTradeseq = sc.read_h5ad(outdir+'/adatas/MultiplexingPreprocessing_unscaled.h5ad')[obs.index]
#Sample extra genes to ensure fair unbyased normalization
ExtraNormGenes = random.sample(adataTradeseq.var_names[~adataTradeseq.var_names.isin(HVGs)].tolist(), len(HVGs.tolist()))
adataTradeseq = adataTradeseq[:,HVGs.tolist()+ExtraNormGenes]

adataTradeseq.obs = obs.loc[adataTradeseq.obs_names]
adataTradeseq = adataTradeseq[:,~adataTradeseq.var_names.str.startswith('MT-')]
adataTradeseq = adataTradeseq[:,~adataTradeseq.var_names.str.startswith(("RPS","RPL"))]
adataTradeseq.obsm = obsm
adataTradeseq.uns["leidenAnnotated_colors"] =  adataBranch[adataTradeseq.obs_names].uns["leidenAnnotated_colors"]


#Trim extreme values of DPT (for potentially low amount of cells)
trimExtremes = [0.01,.99]
minDPT, maxDPT = np.quantile(adataTradeseq.obs["dpt_pseudotime"], trimExtremes)
adataTradeseq = adataTradeseq[(adataTradeseq.obs["dpt_pseudotime"] > minDPT) & (adataTradeseq.obs["dpt_pseudotime"] < maxDPT)]
In [26]:
sc.pl.diffmap(adataTradeseq, color = ["dpt_pseudotime","leidenAnnotated"],
           size = rcParamsDict["dotSize"],add_outline=True, outline_width=(0.2, 0.05), vmin='p2',vmax='p98', frameon=False, wspace=.3, legend_loc="on data")
In [27]:
adataTradeseq.obs[endpoint] = 1
In [28]:
cellWeights = adataTradeseq.obs[endpoint]
cellWeights = np.matrix(cellWeights.to_numpy()).T


pseudotime = adataTradeseq.obs["dpt_pseudotime"]
pseudotime = np.matrix(pseudotime.to_numpy()).T



# Transpose needed!
counts = adataTradeseq.X.T.todense()
counts = counts + abs(counts.min())

varnames = adataTradeseq.var_names.tolist()
In [29]:
import anndata2ri
import rpy2
anndata2ri.activate()
%load_ext rpy2.ipython

The rpy2.ipython extension is already loaded. To reload it, use:
  %reload_ext rpy2.ipython
In [30]:
%%R -i tradeSeqUtilsPath

source(tradeSeqUtilsPath)
In [31]:
%%R -i cellWeights -i pseudotime -i counts -i varnames
set.seed(8)

library(tradeSeq)
library(ggplot2)


w <- as.matrix(cellWeights)
dpt <- as.matrix(pseudotime)
cMatrix <- as.matrix(counts)


gamObj <- fitGAM(cMatrix, verbose = TRUE, pseudotime = dpt, cellWeights = w, nknots = 8, sce=FALSE)
names(gamObj) <- varnames

  |                                                  | 0 % ~calculating   |+                                                 | 1 % ~06m 48s       |++                                                | 2 % ~07m 17s       |++                                                | 3 % ~06m 48s       |+++                                               | 4 % ~06m 42s       |+++                                               | 5 % ~06m 32s       |++++                                              | 6 % ~06m 24s       |++++                                              | 7 % ~06m 24s       |+++++                                             | 8 % ~06m 13s       |+++++                                             | 9 % ~06m 09s       |++++++                                            | 10% ~06m 05s       |++++++                                            | 11% ~06m 02s       |+++++++                                           | 12% ~05m 53s       |+++++++                                           | 13% ~05m 47s       |++++++++                                          | 14% ~05m 43s       |++++++++                                          | 15% ~05m 39s       |+++++++++                                         | 16% ~05m 33s       |+++++++++                                         | 17% ~05m 29s       |++++++++++                                        | 18% ~05m 25s       |++++++++++                                        | 19% ~05m 23s       |+++++++++++                                       | 20% ~05m 18s       |+++++++++++                                       | 21% ~05m 14s       |++++++++++++                                      | 22% ~05m 09s       |++++++++++++                                      | 23% ~05m 04s       |+++++++++++++                                     | 24% ~04m 59s       |+++++++++++++                                     | 25% ~04m 55s       |++++++++++++++                                    | 26% ~04m 53s       |++++++++++++++                                    | 27% ~04m 48s       |+++++++++++++++                                   | 28% ~04m 45s       |+++++++++++++++                                   | 29% ~04m 41s       |++++++++++++++++                                  | 30% ~04m 37s       |++++++++++++++++                                  | 31% ~04m 33s       |+++++++++++++++++                                 | 32% ~04m 27s       |+++++++++++++++++                                 | 33% ~04m 23s       |++++++++++++++++++                                | 34% ~04m 19s       |++++++++++++++++++                                | 35% ~04m 15s       |+++++++++++++++++++                               | 36% ~04m 11s       |+++++++++++++++++++                               | 37% ~04m 07s       |++++++++++++++++++++                              | 38% ~04m 02s       |++++++++++++++++++++                              | 39% ~03m 59s       |+++++++++++++++++++++                             | 40% ~03m 56s       |+++++++++++++++++++++                             | 41% ~03m 52s       |++++++++++++++++++++++                            | 42% ~03m 48s       |++++++++++++++++++++++                            | 43% ~03m 44s       |+++++++++++++++++++++++                           | 44% ~03m 39s       |+++++++++++++++++++++++                           | 45% ~03m 35s       |++++++++++++++++++++++++                          | 46% ~03m 31s       |++++++++++++++++++++++++                          | 47% ~03m 27s       |+++++++++++++++++++++++++                         | 48% ~03m 23s       |+++++++++++++++++++++++++                         | 49% ~03m 20s       |++++++++++++++++++++++++++                        | 51% ~03m 37s       |++++++++++++++++++++++++++                        | 52% ~03m 39s       |+++++++++++++++++++++++++++                       | 53% ~03m 42s       |+++++++++++++++++++++++++++                       | 54% ~03m 44s       |++++++++++++++++++++++++++++                      | 55% ~03m 49s       |++++++++++++++++++++++++++++                      | 56% ~04m 02s       |+++++++++++++++++++++++++++++                     | 57% ~04m 09s       |+++++++++++++++++++++++++++++                     | 58% ~04m 07s       |++++++++++++++++++++++++++++++                    | 59% ~04m 21s       |++++++++++++++++++++++++++++++                    | 60% ~04m 20s       |+++++++++++++++++++++++++++++++                   | 61% ~04m 21s       |+++++++++++++++++++++++++++++++                   | 62% ~04m 23s       |++++++++++++++++++++++++++++++++                  | 63% ~04m 20s       |++++++++++++++++++++++++++++++++                  | 64% ~04m 19s       |+++++++++++++++++++++++++++++++++                 | 65% ~04m 15s       |+++++++++++++++++++++++++++++++++                 | 66% ~04m 21s       |++++++++++++++++++++++++++++++++++                | 67% ~04m 15s       |++++++++++++++++++++++++++++++++++                | 68% ~04m 15s       |+++++++++++++++++++++++++++++++++++               | 69% ~04m 13s       |+++++++++++++++++++++++++++++++++++               | 70% ~04m 18s       |++++++++++++++++++++++++++++++++++++              | 71% ~04m 21s       |++++++++++++++++++++++++++++++++++++              | 72% ~04m 13s       |+++++++++++++++++++++++++++++++++++++             | 73% ~04m 11s       |+++++++++++++++++++++++++++++++++++++             | 74% ~04m 05s       |++++++++++++++++++++++++++++++++++++++            | 75% ~03m 57s       |++++++++++++++++++++++++++++++++++++++            | 76% ~03m 52s       |+++++++++++++++++++++++++++++++++++++++           | 77% ~03m 44s       |+++++++++++++++++++++++++++++++++++++++           | 78% ~03m 34s       |++++++++++++++++++++++++++++++++++++++++          | 79% ~03m 29s       |++++++++++++++++++++++++++++++++++++++++          | 80% ~03m 20s       |+++++++++++++++++++++++++++++++++++++++++         | 81% ~03m 12s       |+++++++++++++++++++++++++++++++++++++++++         | 82% ~03m 07s       |++++++++++++++++++++++++++++++++++++++++++        | 83% ~02m 60s       |++++++++++++++++++++++++++++++++++++++++++        | 84% ~02m 49s       |+++++++++++++++++++++++++++++++++++++++++++       | 85% ~02m 39s       |+++++++++++++++++++++++++++++++++++++++++++       | 86% ~02m 29s       |++++++++++++++++++++++++++++++++++++++++++++      | 87% ~02m 19s       |++++++++++++++++++++++++++++++++++++++++++++      | 88% ~02m 12s       |+++++++++++++++++++++++++++++++++++++++++++++     | 89% ~02m 03s       |+++++++++++++++++++++++++++++++++++++++++++++     | 90% ~01m 52s       |++++++++++++++++++++++++++++++++++++++++++++++    | 91% ~01m 41s       |++++++++++++++++++++++++++++++++++++++++++++++    | 92% ~01m 31s       |+++++++++++++++++++++++++++++++++++++++++++++++   | 93% ~01m 20s       |+++++++++++++++++++++++++++++++++++++++++++++++   | 94% ~01m 09s       |++++++++++++++++++++++++++++++++++++++++++++++++  | 95% ~58s           |++++++++++++++++++++++++++++++++++++++++++++++++  | 96% ~46s           |+++++++++++++++++++++++++++++++++++++++++++++++++ | 97% ~35s           |+++++++++++++++++++++++++++++++++++++++++++++++++ | 98% ~23s           |++++++++++++++++++++++++++++++++++++++++++++++++++| 99% ~12s           |++++++++++++++++++++++++++++++++++++++++++++++++++| 100% elapsed=19m 28s
In [32]:
%%R -i tradeSeqUtilsPath

source(tradeSeqUtilsPath)
In [33]:
dptseq = np.linspace(pseudotime.min(),pseudotime.max(), 5)
nGenes = 3

leidenCols = dict(zip(adataTradeseq.obs.leidenAnnotated.cat.categories, adataTradeseq.uns["leidenAnnotated_colors"]))
In [34]:
%%R -o assoRes  -o startRes -i minDPT -i maxDPT


startRes <- startVsEndTest(gamObj, pseudotimeValues = c(minDPT+.01,maxDPT-.01))
In [35]:
# Top Up and Down
startResFilt = startRes[(startRes.pvalue <= 0.001) & (abs(startRes.logFClineage1) >= 2)]
startResFilt["Gene"] = startResFilt.index.tolist()
startResFilt["test"] = "startVsEndTest"
startResGenesPositive = startResFilt[startResFilt.logFClineage1 < 0 ].sort_values(["pvalue","logFClineage1"], ascending=[True,True])
startResGenesPositive_plot = startResGenesPositive.head(nGenes).index.tolist()
startResGenesPositive["PatternType"] = "decreasing"
startResGenesPositive = startResGenesPositive.rename(columns={"logFClineage1":"logFC"})

startResGenesNegatives = startResFilt[startResFilt.logFClineage1 > 0 ].sort_values(["pvalue","logFClineage1"], ascending=[True,False])
startResGenesNegatives_plot = startResGenesNegatives.head(nGenes).index.tolist()
startResGenesNegatives["PatternType"] = "increasing"
startResGenesNegatives = startResGenesNegatives.rename(columns={"logFClineage1":"logFC"})
In [37]:
CombinedDF = pd.concat([startResGenesPositive,startResGenesNegatives], ignore_index=True)
CombinedDF["Branch"] = branchTag
CombinedDF.to_csv(outdir+"/Tradeseq."+branchTag+".tsv", sep="\t", index=False)
In [39]:
%%R -i startResGenesPositive_plot -i startResGenesNegatives_plot -o assoResSSPandas

CombinedList <- c(startResGenesPositive_plot,startResGenesNegatives_plot)

datalist <- list()
for (g in CombinedList){
    
    localModel <- gamObj[[g]]    
    data <- localModel$model    
    y <- data$y
    nCurves <- length(localModel$smooth)

    for (jj in seq_len(nCurves)) {
        df <- .getPredictRangeDf(localModel$model, jj, nPoints = 100)
        yhat <- predict(localModel, newdata = df, type = "response")

        Newframe <- data.frame("fittedCounts" = yhat)
        colnames(Newframe) <- c(paste0("fittedCounts.",g))

        datalist[[paste(g,jj)]] <- Newframe

        
        

    }

}

assoResSSPandas =  do.call(cbind, datalist)
In [40]:
#Curatedmarkerslist = CuratedmarkersDF.loc[CuratedmarkersDF.score > 1,"name"].tolist()
#CuratedmarkersDF = pd.read_csv(markers, header=None, sep = "\t", names=["name","area","score"])
#Curatedmarkerslist = CuratedmarkersDF.loc[CuratedmarkersDF.score >= 1,"name"].tolist()

##############################


###################à
plt.rcdefaults()

sns.set_style("ticks", {'axes.grid' : False})


fig, axes = plt.subplots(2,1, figsize=(12,5),sharex=True, 
                         gridspec_kw={'height_ratios': [1, 3]})

sns.boxplot(data=adataTradeseq.obs, x="dpt_pseudotime", y="leidenAnnotated", ax = axes[0],
            palette=leidenCols,
            showfliers=False,order=leidenOrder)

    

    
    


#fig, axes = plt.subplots(2,1, figsize=(10,5),sharex=True, 
#                         gridspec_kw={'height_ratios': [1, 2]})

#sns.histplot(data=adataTradeseq.obs, x="dpt_pseudotime", hue="leidenAnnotated", ax=axes[0])

# sns.histplot(data=adataTradeseq.obs, x="dpt_pseudotime", hue="leidenAnnotated", ax=axes[0],bins=100,
#               palette=dict(zip(adata.obs.leidenAnnotated.cat.categories, adata.uns["leidenAnnotated_colors"])), alpha=.5)




SmoothedExpr = assoResSSPandas.copy()
SmoothedExpr["pseudotime"] = SmoothedExpr.index.tolist()      
SmoothedExpr["pseudotime"] =  np.linspace(pseudotime.min(),pseudotime.max(), 100)
SmoothedExpr = SmoothedExpr.melt("pseudotime")
SmoothedExpr["value"] = np.log(SmoothedExpr["value"])


#TrimExtremes


sns.lineplot(data=SmoothedExpr, x="pseudotime", y="value", hue= "variable",  linewidth=3, ax=axes[1])
#axes[1].lines[1].set_linestyle("--")





dptAxis = np.linspace(SmoothedExpr.pseudotime.min(), SmoothedExpr.pseudotime.max(), 10).round(2)      
ylimits = np.array([SmoothedExpr.value.min()-abs((SmoothedExpr.value.min()*.1)), 
                    SmoothedExpr.value.max()+abs((SmoothedExpr.value.max()*.1))]).round(2)
exprAxis = np.linspace(ylimits.min(), ylimits.max(),5).round(2)


axes[1].set_xlim([dptAxis.min(), dptAxis.max()])
axes[1].set_ylim(ylimits)


axes[0].yaxis.tick_right()
axes[0].spines[['left', 'top','bottom']].set_visible(False)
axes[0].spines['right'].set_linewidth(1.5)
axes[0].tick_params(axis='y', which='major', labelsize=15, width=1.5, length=10)
axes[0].tick_params(axis='x', which='major', labelsize=15, width=0, length=10)


axes[1].set_xticks(dptAxis)
axes[1].set_yticks(exprAxis)

axes[1].spines['left'].set_linewidth(1.5)
axes[1].spines['bottom'].set_linewidth(1.5)
axes[1].tick_params(axis='both', which='major', labelsize=15, width=1.5, length=10)
axes[1].spines[['right', 'top']].set_visible(False)
axes[1].xaxis.tick_bottom()
axes[1].yaxis.tick_left()
axes[1].legend(title='Gene', bbox_to_anchor=(1, .7))


sns.despine(ax=axes[1], top=True, right=True, left=False, bottom=False, offset=10, trim=False)
sns.despine(ax=axes[0], top=True, right=False, left=False, bottom=True, offset=10, trim=False)
axes[0].yaxis.set_label_position("right")
axes[0].spines[['left', 'top','bottom']].set_visible(False)

axes[1].set(xlabel='dpt pseudotime', ylabel='Log smoothed expression')



fig.show()
        


fig.savefig(figDir+'/TopGenes.'+branchTag+'.pdf',bbox_inches='tight')
In [ ]: